The rate of the Atlantic Meridional overturning circulation's (AMOC) weakening response to CO₂ increase is influenced by the anomalous surface fluxes directly induced by the CO₂ increase, as well as several feedbacks triggered by the AMOC's response itself. The relative roles of feedbacks on the AMOC's response to CO₂ increase remains unclear. This work explores the relative roles of surface flux feedbacks and oceanic temperature- and salinity- advection feedbacks using a suite of partially coupled experiments and tracer decomposition in the CESM1.2 model. The tracer decomposition method isolates the circulation-driven from surface-driven ocean temperature and salinity anomalies, enabling the quantification of the impact of ocean temperature and salinity advection feedbacks. The impact of the feedbacks due to the individual component of the surface flux response to the AMOC's weakening are also accessed using partially coupled experiments where the surface flux response to the AMOC's weakening is suppressed. The results show that while CO₂-induced anomalous surface heat fluxes primarily drive an initial weakening of the AMOC, the impact of feedbacks on the AMOC kick in after 15 years, causing a faster AMOC weakening rate, and a net AMOC weakening that is about twice the weakening directly induced by the CO₂ increase. The AMOC's weakening is mainly enhanced by a positive surface heat flux feedback on the AMOC's weakening response. This positive feedback works by neutralizing the negative temperature advection feedback, so that the impact of the positive salinity advection feedback on the AMOC's response becomes dominant. In the partially coupled experiment where this surface heat flux feedback is suppressed, ocean temperature advection feedback halts the AMOC weakening after 20 years. The momentum and freshwater responses to the AMOC provide negligible feedback on the AMOC. Similarly, anomalous freshwater fluxes from Arctic sea ice melt play a minimal role in the AMOC's response due to a propagation path that prevent these fresh salt anomalies from reaching the North Atlantic deep water formation regions. These results emphasize the importance of reducing salinity biases in GCMs for accurate projection of the AMOC changes and stability.